Microsoft API Pricing (Updated 2025)
This page tracks Microsoft API pricing for all 8 models. Prices are shown per 1,000 tokens (cost per token) with clear examples so you can estimate spend quickly.
Current Prices (per 1,000 tokens)
Model | Input ($/1K tokens) | Output ($/1K tokens) | Context Length |
|---|---|---|---|
mai-ds-r1 | $0.000200 | $0.000800 | 163,840 |
phi-3-medium-128k-instruct | $0.001000 | $0.001000 | 128,000 |
phi-3-mini-128k-instruct | $0.000100 | $0.000100 | 128,000 |
phi-3.5-mini-128k-instruct | $0.000100 | $0.000100 | 128,000 |
phi-4 | $0.000060 | $0.000140 | 16,384 |
phi-4-multimodal-instruct | $0.000050 | $0.000100 | 131,072 |
phi-4-reasoning-plus | $0.000070 | $0.000350 | 32,768 |
wizardlm-2-8x22b | $0.000480 | $0.000480 | 65,536 |
* Some models use tiered pricing based on prompt length. Displayed prices are for prompts ≤ 200k tokens.
What does "cost per token" mean?
Tokens are the basic units that AI models use to process text. Generally, 1,000 tokens ≈ ~750 words. The cost per token determines how much you pay for each unit of text processed by the model.
Formula: Total Cost = (tokens / 1,000) × price per 1K tokens
Examples
Example usage: A 500-word prompt + 300-word response ≈ ~1,067 tokens total.
- • 1,000 tokens ≈ ~750 words of text
- • Short email: ~200-400 tokens
- • Blog post: ~1,000-3,000 tokens
- • Research paper: ~10,000+ tokens
Compare with Other Providers
See how Microsoft compares with other AI providers including OpenAI, Anthropic, Google, and more.
View full pricing comparison →